about mean and std

[hanguniverse]

Thanks for your wonderful work, I have a question that in ImageReward.py l80-81 set the self.mean=0.167 and self.std=1.033,
How did you get these two numbers？

[xujz18]

Thanks for your attention! You can refer #7

[hanguniverse]

@xujz18 thanks for your reply! I have another question that when I evaluate the reward model on a new dataset, I want to let the reward score in [0,1], How can I do that？ thank you！

[xujz18]

Very happy to discuss this with you. I think you can get the scores for the whole dataset after normalizing them based on the mean and variance of the new dataset.